This is an automated email from the ASF dual-hosted git repository.
sergeykamov pushed a commit to branch NLPCRAFT-491
in repository https://gitbox.apache.org/repos/asf/incubator-nlpcraft.git
commit aafd349afd82fe4aeae7f132a8fad8e9e9ea17fe
Author: Sergey Kamov <skhdlem...@gmail.com>
AuthorDate: Fri Apr 8 11:51:04 2022 +0300
WIP.
---
.../pizzeria/components/ElementExtender.scala | 78 +++++++++++++---------
1 file changed, 48 insertions(+), 30 deletions(-)
diff --git
a/nlpcraft-examples/pizzeria/src/main/java/org/apache/nlpcraft/examples/pizzeria/components/ElementExtender.scala
b/nlpcraft-examples/pizzeria/src/main/java/org/apache/nlpcraft/examples/pizzeria/components/ElementExtender.scala
index 364f1058..69f4a409 100644
---
a/nlpcraft-examples/pizzeria/src/main/java/org/apache/nlpcraft/examples/pizzeria/components/ElementExtender.scala
+++
b/nlpcraft-examples/pizzeria/src/main/java/org/apache/nlpcraft/examples/pizzeria/components/ElementExtender.scala
@@ -23,6 +23,7 @@ import java.util
import java.util.List as JList
import scala.collection.mutable
import scala.jdk.CollectionConverters.*
+import com.typesafe.scalalogging.LazyLogging
/**
*
@@ -32,39 +33,56 @@ import scala.jdk.CollectionConverters.*
case class EntityData(id: String, property: String)
/**
- * Simple element extender.
- * For each 'main' element it tries to find extra element. If it cane done
instead of this pair new element created.
- * This new element has 'main' element ID. Extra element configured property
copied to this new element.
+ * Element extender.
+ * For each 'main' element it tries to find related extra element and to make
new complex element instead of this pair.
+ * This new element has:
+ * 1. Same ID as main element, also all main element properties copied into
this new element's properties.
+ * 2. Tokens from both elements.
+ * 3. Configured property from extra element copied into new element's
properties.
*
- * Note that for simplified implementation, it doesn't care about words
between pairs elements and pairs,
- * also it doesn't check correctness of words order.
+ * Note that it is simple example implementation.
+ * It just tries to unite nearest neighbours and doesn't check intermediate
words, order correctness etc.
*/
-case class ElementExtender(mainDataSeq: Seq[EntityData], extraData:
EntityData) extends NCEntityMapper:
- private def getToks(e: NCEntity): mutable.Seq[NCToken] =
e.getTokens.asScala
- override def map(req: NCRequest, cfg: NCModelConfig, entities:
util.List[NCEntity]): util.List[NCEntity] =
- val mainDataMap = mainDataSeq.map(p => p.id -> p).toMap
+object ElementExtender:
+ case class EntityHolder(element: NCEntity):
+ lazy val position: Double =
+ val toks = element.getTokens.asScala
+ (toks.head.getIndex + toks.last.getIndex) / 2.0
- var es = entities.asScala
- val main = es.filter(e => mainDataMap.contains(e.getId))
- val extra = es.filter(_.getId == extraData.id)
+ private def toTokens(e: NCEntity): mutable.Seq[NCToken] =
e.getTokens.asScala
- if main.nonEmpty && main.size >= extra.size then
- var ok = true
- val mapped =
- for ((e1, e2) <- main.zip(extra) if ok) yield
- if e1.getId == e2.getId then
- ok = false
- null
- else
- val (mEnt, eEnt) = if mainDataMap.contains(e1.getId)
then (e1, e2) else (e2, e1)
- new NCPropertyMapAdapter with NCEntity:
- mEnt.keysSet().forEach(k => put(k, mEnt.get(k)))
- put[String](mainDataMap(mEnt.getId).property,
eEnt.get[String](extraData.property).toLowerCase)
- override val getTokens: JList[NCToken] =
(getToks(mEnt) ++ getToks(eEnt)).sortBy(_.getIndex).asJava
- override val getRequestId: String =
req.getRequestId
- override val getId: String = mEnt.getId
+import ElementExtender.*
- es = es --= main.take(mapped.size)
- es = es --= extra.take(mapped.size)
- (es ++ mapped).sortBy(getToks(_).head.getIndex).asJava
+/**
+ *
+ * @param mainSeq
+ * @param extra
+ */
+case class ElementExtender(mainSeq: Seq[EntityData], extra: EntityData)
extends NCEntityMapper with LazyLogging:
+ override def map(req: NCRequest, cfg: NCModelConfig, entities:
JList[NCEntity]): JList[NCEntity] =
+ def combine(m: NCEntity, mProp: String, e: NCEntity): NCEntity =
+ new NCPropertyMapAdapter with NCEntity:
+ m.keysSet().forEach(k => put(k, m.get(k)))
+ put[String](mProp, e.get[String](extra.property).toLowerCase)
+ override val getTokens: JList[NCToken] = (toTokens(m) ++
toTokens(e)).sortBy(_.getIndex).asJava
+ override val getRequestId: String = req.getRequestId
+ override val getId: String = m.getId
+
+ val es = entities.asScala
+ val mainById = mainSeq.map(p => p.id -> p).toMap
+ val mainHs = mutable.HashSet.empty ++ es.filter(e =>
mainById.contains(e.getId)).map(p => EntityHolder(p))
+ val extraHs = es.filter(_.getId == extra.id).map(p => EntityHolder(p))
+
+ if mainHs.nonEmpty && mainHs.size >= extraHs.size then
+ val main2Extra = mutable.HashMap.empty[NCEntity, NCEntity]
+
+ for (e <- extraHs)
+ val m = mainHs.minBy(m => Math.abs(m.position - e.position))
+ mainHs -= m
+ main2Extra += m.element -> e.element
+
+ val newEs = for ((m, e) <- main2Extra) yield combine(m,
mainById(m.getId).property, e)
+ val used = (mainHs.map(_.element) ++ extraHs.map(_.element)).toSet
+
+ (es.filter(e => !used.contains(e)) ++ mainHs.map(_.element) ++
newEs).sortBy(toTokens(_).head.getIndex).asJava
else entities
\ No newline at end of file
